You have some options, of all which are viable primarily for prototypes or engineering evaluations

(1) Re-ball the BGA package with Pb-free solder balls
- can be fairly reliable but is time consuming and not viable for production volumes. There are folks that provide this service on outsource basis if you have higher volumes
(2) Use a BGA rework station to attach the Pb/Sn packages as a secondary operation - again reasonably reliable but not viable for production volumes
(3) Use mini-stencil to apply low temp solder paste to the PCB (if real-estate permits access). Hand place your Pb/Sn BGA packages using 2DX and reflow as secondary operation at lower temperature (175oC). This is tricky since solder bridging is difficult to avoid.

We have been placing SnPb spheres on ENIG for as long as there have been BGA packages .. with good success.

Some of the other mixes are dependant upon the specific finish material employed. IMSn, and IMAg have some concerned.

When reballing, be aware the of alloy created where the replacement Sphere contacts the component attachment as well as the board itself. During reballing some of the solder remains with the component where the previous sphere was attached.

First is the Sn/Pb BGA going to be attached with Sn/Pb on to a lead free finished board such as a lead free HASL or immersions silver? If so this is OK. If this is not the case then reliability of a mixed lead and lead free is questionable. Cooling rates are critical these will affect the lead distribution through out the solder connection, slower is generally not desirable as it allows the lead to segregate in the last to section to cool. How long the Sn/Pb is diffused into the lead free all will case variations in the reliability of the components if lead is well diffused and rapid freezing is controlled there is less of an issue with the lead segregating. This leads to a more desirable solder connection however lead in lead free should try to be eliminated.